From 64b41cc4f637cd562bf54de4e0b78c2acddb3137 Mon Sep 17 00:00:00 2001 From: vincentccc Date: Mon, 8 Sep 2025 22:26:19 +0800 Subject: [PATCH] update --- README.md | 6 ++++++ workspace/lerobot_example/config_qact.yml | 2 ++ 2 files changed, 8 insertions(+) diff --git a/README.md b/README.md index 18d59e5..a3252de 100644 --- a/README.md +++ b/README.md @@ -76,10 +76,16 @@ Training script path configuration - Robot DOF configuration - Training hyperparameters +Download the Flow/FAST pretrained model and run: ```bash bash ./workspace/lerobot_example/run.sh ``` +To fine-tune using the FAST tokenizer, please download the repository and update the `action_tokenizer_path`. Make sure to set `use_fast_tokenizer` to `true`: +```bash +git clone https://huggingface.co/physical-intelligence/fast +``` + ## Inference For model inference, please refer to: diff --git a/workspace/lerobot_example/config_qact.yml b/workspace/lerobot_example/config_qact.yml index fa683ae..b4d2fea 100644 --- a/workspace/lerobot_example/config_qact.yml +++ b/workspace/lerobot_example/config_qact.yml @@ -6,6 +6,8 @@ log_name: "robotic_training" log_project: "vla_training" model_type: qwen2_5 pretrained_qwen_vl_path: "/path/to/qwen_vl_model/" +# whether to enable fast tokenizer +use_fast_tokenizer: false action_tokenizer_path: "/path/to/fast/" save_path: "/path/to/workspace/"